Tax Amnesty Uptake in Light of Behavioral Intervention: A Natural Field Experiment

This is a messaging study that includes behavioural intervention sentences in emails. The emails are sent out by the administration and are similar in content except for the sentence that implies social norm, deterrence or benefit depending on the group a subject is assigned.
Since the subjects receive communication through email, the behavioural intervention in the email content is reflected in the subject line. In the email body, the subject is addressed by name and informed about the amnesty, the period within which it runs, and the arrears accrued by them. The next sentence is the treatment, made salient by bolding. The email concludes with links to the amnesty guidelines and contact details. Apart from the treatment variation depending on random assignment, message content is the same for all the subjects.

Since the subjects receive communication through email, the behavioural intervention in the email content is reflected in the subject line. For example, those who receive a deterrence message have a subject line reading “apply for amnesty now to avoid enforcement”. In the email body, the subject is addressed by name and informed about the amnesty, the period within which it runs, and the principal tax arrears accrued by them. The next sentence is the treatment, made salient by bolding. The email concludes with links to the amnesty guidelines and contact details. Apart from the treatment variation depending on random assignment message content is the same for all the subjects.

2023-12-11

2024-06-30

Amnesty uptake, Arrears payment,

The study will use regression analysis to estimate average treatment effects in the first phase of the experiment. Here we measure how behavioural intervention affects amnesty uptake and subsequent payment.
In the second phase, we measure dynamic treatment effects arising from a treatment assignment path and so estimate a logistic autoregressive model. Here, we evaluate the optimal sequence in amnesty uptake and settlement of arrears.

Data correction

The study evaluates the behavioural intervention that leads to amnesty uptake and subsequent settlement by subjects in arrears through a natural field experiment. It also examines treatment sequencing and its role in influencing compliance with the amnesty. Through randomised controlled trials, 34,886 delinquents are randomly assigned to four arms. In the first phase, they receive benefit-oriented messages, deterrence messages, and dynamic social norms. The control group only receives information about the amnesty. Those in the social norm and benefit treatment arms that do not take up the amnesty by the second round, receive sequenced messages in the second third and fourth round of reminders. The sequence continues for non-takers in the social norm and benefit group until they receive each of the three treatment messages. Following from literature, the deterrence treatment is the strongest signal, we therefore do not apply sequencing of treatment to subjects in this arm. Instead, subjects in this arm are treated with reminders. Outcomes of interest are amnesty uptake and subsequent payments. We extend behavioural intervention literature by focusing on delinquents in an amnesty when policy is silent on heightened enforcement measures post amnesty. In addition, we introduce sequencing in a within subject design and evaluate its effectiveness in enhancing compliance behaviour.

The study applies randomised controlled trials in a natural field experiment. The experiment is divided into two phases. In the first phase, a between subject design is adopted to include behavioural intervention in emails sent by the tax administration to sampled tax delinquents. The sample consists of 34,886 tax delinquents who are divided into four groups. In the control group, we assign 9,275 subjects whereas each treatment group has 8537 subjects. The control group receives general information about the existence of an amnesty. This is to take care of the effect of receiving any message from tax administration regardless of its contents (Perez-Truglia and Troiano, 2018; Mascagni et al., 2017). In addition to the amnesty information, the three treatment groups receive salient messages on amnesty payment plan benefit, enforcement post amnesty and amnesty social norms observed from tax delinquents. Hence treatment implies deterrence, benefit, and dynamic social norms.
Since the subjects receive communication through email, the behavioural intervention in the email content is reflected in the subject line. For example, those who receive a deterrence message have a subject line reading “apply for amnesty now to avoid enforcement”. In the email body, the subject is addressed by name and informed about the amnesty, the period within which it runs, and the principal tax arrears accrued by them. The next sentence is the treatment, made salient by bolding. The email concludes with links to the amnesty guidelines and contact details. Apart from the treatment variation depending on random assignment message content is the same for all the subjects.
Looking at the dynamic social norm message, defined by Wenzel (2004) as prevalence or acceptance of a given behaviour among a reference group, compliance is driven by the choice of others in the group. If a taxpayer believes that compliance is high in his specific reference group, the likelihood of compliance is high especially if they identify with the group. The design refers to other taxpayers with tax arrears as a reference group against which a taxpayer analyses his/her decision to comply. Studies find evidence that social norms have an important role in establishing tax compliance (Hallworth et al., 2017; Del Carpio, 2013; Kirchler et al. 2014). We therefore inform the subjects that more taxpayers are taking up the amnesty and encourage them to follow suit. In doing this, we expect taxpayers to respond by considering what their cohorts are doing.
Unlike vast majority of literature where deterrence messages carry a tough tone, we apply a soft enforcement message. Tax delinquents are reminded that at the end of the amnesty, tax debt including penalties and interest is payable and will be enforced as provided for in the tax law. This kind of messaging has been found to be effective in enhancing compliance (Fellner et al., 2013; Gil et al., 2023; Holz et al., 2023; Perez-Truglia and Troiano, 2018; Castro and Scartascini,2015). However, other studies find a backfiring effect from enforcement messages, especially from high income earners (Mascagni and Nell, 2021; Slemrod et al., 1999). In our case, we expect the deterrence message to have a positive effect on tax amnesty uptake and subsequent payment.
In the third treatment, we use a gain framed message. Besides extending the benefit of a payment plan to the taxpayer, the message points out the positive psychological benefits that come from being compliant. Loewenstein et al. (2001) indicates that feelings can drive decision making behaviour. We assume the tax debt and frequent demand letters create distress to tax delinquents and they would be motivated to take up the amnesty and pay at their pace or correct their tax ledger in exchange for peace of mind.
In the second phase of the experiment, we follow a treatment assignment path by sequencing message treatments culminating to a deterrence message. In the sequencing phase, each of the three treatments are broken down into three subgroups. For instance, those yet to take up the amnesty in the social norm group are divided into three subgroups. The first subgroup is observed in the third and fourth round but receives a deterrence message in the fifth round. The fifth round is expected to be administered towards the end of the amnesty and therefore captures time effect occasioned by last minute rush. It is only this group that goes to the fifth round. The second subgroup is observed in the third round and treated with a deterrence message in the fourth and final round. The third subgroup goes through the sequence and receives a benefit message in the third round followed by a deterrence message in the fourth and final round. The decision to have deterrence as a last treatment is motivated by finding from literature which demonstrates deterrence to be the strongest signal.
Although the deterrence group has three subgroups, no sequencing of treatments is done. Moreover, in the context of the slippery slope framework (Kirchler,2007), a down grade to softer tax morale messages would imply a lack of power to enforce by tax authorities hence reduced trust towards tax administration. It is for this reason that we also retain deterrence treatment within the group to the end. Like the other treatment groups, one subgroup is observed in the third and fourth round but only receives a reminder in the fifth round. The second group is observed in the third round but treated with a reminder in the fourth and final round. Lastly, the third subgroup receives a reminder in the third and fourth round.
By implementing this design in the second phase, we intend to tease out effects that maybe associated with the spontaneous decision makers and those who take time to make compliance decisions. At the same time, it is possible to decipher the sequence of message interventions that are more effective in driving compliance choices. As the treatment culminates to deterrence, we can observe whether compliance is driven by the message or lapse of time in the fifth round. In all rounds, we retain a consistent control group by sending reminders in each round of the treatment.

In office by computer

Individual

No

Four groups to include tree treatments and one control from one of six administrative regions.

34,886 individuals

Control 9275
Benefit 8537
Deterrence 8537
Social Norm 8537

Minimum detectable effect ≈ 3%, power ≈ 80% sample size 34,886 individuals/firms standard deviation ≈ 6.67